3D U2-Net: A 3D Universal U-Net for Multi-domain Medical Image Segmentation

3D U2-Net: A 3D Universal U-Net for Multi-domain Medical Image Segmentation. Huang, C., Han, H., Yao, Q., Zhu, S., & Zhou, K., S. Volume 1 , Springer International Publishing, 2019.

Website doi abstract bibtex 1 download

Fully convolutional neural networks like U-Net have been the state-of-the-art methods in medical image segmentation. Practically, a network is highly specialized and trained separately for each segmenta- tion task. Instead of a collection of multiple models, it is highly desirable to learn a universal data representation for different tasks, ideally a sin- gle model with the addition of a minimal number of parameters steered to each task. Inspired by the recent success of multi-domain learning in image classification, for the first time we explore a promising univer- sal architecture that handles multiple medical segmentation tasks and is extendable for new tasks, regardless of different organs and imaging modalities. Our 3D Universal U-Net (3D U2-Net) is built upon sepa- rable convolution, assuming that images from different domains have domain-specific spatial correlations which can be probed with channel- wise convolution while also share cross-channel correlations which can be modeled with pointwise convolution. We evaluate the 3D U2-Net on five organ segmentation datasets. Experimental results show that this uni- versal network is capable of competing with traditional models in terms of segmentation accuracy, while requiring only about 1% of the param- eters. Additionally, we observe that the architecture can be easily and effectively adapted to a new domain without sacrificing performance in the domains used to learn the shared parameterization of the universal network. We put the code of 3D U2-Net into public domain (https:// github.com/huangmozhilv/u2net

@book{
 title = {3D U2-Net: A 3D Universal U-Net for Multi-domain Medical Image Segmentation},
 type = {book},
 year = {2019},
 source = {Proceeding of the International Conference on Medical Image Computing and Computer Assisted Interventions},
 keywords = {Multi-domain learning,Segmentation,Universal model},
 pages = {291-299},
 volume = {1},
 websites = {http://dx.doi.org/10.1007/978-3-030-32245-8_33},
 publisher = {Springer International Publishing},
 id = {dc353a48-a502-307f-a646-8928f5bbee35},
 created = {2024-01-13T06:15:56.027Z},
 file_attached = {true},
 profile_id = {d5b9f595-4104-33d2-b640-e84dee1b7ded},
 last_modified = {2024-01-13T06:24:48.304Z},
 read = {false},
 starred = {false},
 authored = {true},
 confirmed = {true},
 hidden = {false},
 private_publication = {false},
 abstract = {Fully convolutional neural networks like U-Net have been the state-of-the-art methods in medical image segmentation. Practically, a network is highly specialized and trained separately for each segmenta- tion task. Instead of a collection of multiple models, it is highly desirable to learn a universal data representation for different tasks, ideally a sin- gle model with the addition of a minimal number of parameters steered to each task. Inspired by the recent success of multi-domain learning in image classification, for the first time we explore a promising univer- sal architecture that handles multiple medical segmentation tasks and is extendable for new tasks, regardless of different organs and imaging modalities. Our 3D Universal U-Net (3D U2-Net) is built upon sepa- rable convolution, assuming that images from different domains have domain-specific spatial correlations which can be probed with channel- wise convolution while also share cross-channel correlations which can be modeled with pointwise convolution. We evaluate the 3D U2-Net on five organ segmentation datasets. Experimental results show that this uni- versal network is capable of competing with traditional models in terms of segmentation accuracy, while requiring only about 1% of the param- eters. Additionally, we observe that the architecture can be easily and effectively adapted to a new domain without sacrificing performance in the domains used to learn the shared parameterization of the universal network. We put the code of 3D U2-Net into public domain (https:// github.com/huangmozhilv/u2net},
 bibtype = {book},
 author = {Huang, Chao and Han, Hu and Yao, Qingsong and Zhu, Shankan and Zhou, Kevin S.},
 doi = {10.1007/978-3-030-32245-8}
}

Downloads: 1

{"_id":"4p8QRNQvoDFkriaAi","bibbaseid":"huang-han-yao-zhu-zhou-3du2neta3duniversalunetformultidomainmedicalimagesegmentation-2019","author_short":["Huang, C.","Han, H.","Yao, Q.","Zhu, S.","Zhou, K., S."],"bibdata":{"title":"3D U2-Net: A 3D Universal U-Net for Multi-domain Medical Image Segmentation","type":"book","year":"2019","source":"Proceeding of the International Conference on Medical Image Computing and Computer Assisted Interventions","keywords":"Multi-domain learning,Segmentation,Universal model","pages":"291-299","volume":"1","websites":"http://dx.doi.org/10.1007/978-3-030-32245-8_33","publisher":"Springer International Publishing","id":"dc353a48-a502-307f-a646-8928f5bbee35","created":"2024-01-13T06:15:56.027Z","file_attached":"true","profile_id":"d5b9f595-4104-33d2-b640-e84dee1b7ded","last_modified":"2024-01-13T06:24:48.304Z","read":false,"starred":false,"authored":"true","confirmed":"true","hidden":false,"private_publication":false,"abstract":"Fully convolutional neural networks like U-Net have been the state-of-the-art methods in medical image segmentation. Practically, a network is highly specialized and trained separately for each segmenta- tion task. Instead of a collection of multiple models, it is highly desirable to learn a universal data representation for different tasks, ideally a sin- gle model with the addition of a minimal number of parameters steered to each task. Inspired by the recent success of multi-domain learning in image classification, for the first time we explore a promising univer- sal architecture that handles multiple medical segmentation tasks and is extendable for new tasks, regardless of different organs and imaging modalities. Our 3D Universal U-Net (3D U2-Net) is built upon sepa- rable convolution, assuming that images from different domains have domain-specific spatial correlations which can be probed with channel- wise convolution while also share cross-channel correlations which can be modeled with pointwise convolution. We evaluate the 3D U2-Net on five organ segmentation datasets. Experimental results show that this uni- versal network is capable of competing with traditional models in terms of segmentation accuracy, while requiring only about 1% of the param- eters. Additionally, we observe that the architecture can be easily and effectively adapted to a new domain without sacrificing performance in the domains used to learn the shared parameterization of the universal network. We put the code of 3D U2-Net into public domain (https:// github.com/huangmozhilv/u2net","bibtype":"book","author":"Huang, Chao and Han, Hu and Yao, Qingsong and Zhu, Shankan and Zhou, Kevin S.","doi":"10.1007/978-3-030-32245-8","bibtex":"@book{\n title = {3D U2-Net: A 3D Universal U-Net for Multi-domain Medical Image Segmentation},\n type = {book},\n year = {2019},\n source = {Proceeding of the International Conference on Medical Image Computing and Computer Assisted Interventions},\n keywords = {Multi-domain learning,Segmentation,Universal model},\n pages = {291-299},\n volume = {1},\n websites = {http://dx.doi.org/10.1007/978-3-030-32245-8_33},\n publisher = {Springer International Publishing},\n id = {dc353a48-a502-307f-a646-8928f5bbee35},\n created = {2024-01-13T06:15:56.027Z},\n file_attached = {true},\n profile_id = {d5b9f595-4104-33d2-b640-e84dee1b7ded},\n last_modified = {2024-01-13T06:24:48.304Z},\n read = {false},\n starred = {false},\n authored = {true},\n confirmed = {true},\n hidden = {false},\n private_publication = {false},\n abstract = {Fully convolutional neural networks like U-Net have been the state-of-the-art methods in medical image segmentation. Practically, a network is highly specialized and trained separately for each segmenta- tion task. Instead of a collection of multiple models, it is highly desirable to learn a universal data representation for different tasks, ideally a sin- gle model with the addition of a minimal number of parameters steered to each task. Inspired by the recent success of multi-domain learning in image classification, for the first time we explore a promising univer- sal architecture that handles multiple medical segmentation tasks and is extendable for new tasks, regardless of different organs and imaging modalities. Our 3D Universal U-Net (3D U2-Net) is built upon sepa- rable convolution, assuming that images from different domains have domain-specific spatial correlations which can be probed with channel- wise convolution while also share cross-channel correlations which can be modeled with pointwise convolution. We evaluate the 3D U2-Net on five organ segmentation datasets. Experimental results show that this uni- versal network is capable of competing with traditional models in terms of segmentation accuracy, while requiring only about 1% of the param- eters. Additionally, we observe that the architecture can be easily and effectively adapted to a new domain without sacrificing performance in the domains used to learn the shared parameterization of the universal network. We put the code of 3D U2-Net into public domain (https:// github.com/huangmozhilv/u2net},\n bibtype = {book},\n author = {Huang, Chao and Han, Hu and Yao, Qingsong and Zhu, Shankan and Zhou, Kevin S.},\n doi = {10.1007/978-3-030-32245-8}\n}","author_short":["Huang, C.","Han, H.","Yao, Q.","Zhu, S.","Zhou, K., S."],"urls":{"Website":"http://dx.doi.org/10.1007/978-3-030-32245-8_33"},"biburl":"https://bibbase.org/service/mendeley/d5b9f595-4104-33d2-b640-e84dee1b7ded","bibbaseid":"huang-han-yao-zhu-zhou-3du2neta3duniversalunetformultidomainmedicalimagesegmentation-2019","role":"author","keyword":["Multi-domain learning","Segmentation","Universal model"],"metadata":{"authorlinks":{}},"downloads":1},"bibtype":"book","biburl":"https://bibbase.org/service/mendeley/d5b9f595-4104-33d2-b640-e84dee1b7ded","dataSources":["ya2CyA73rpZseyrZ8","2252seNhipfTmjEBQ"],"keywords":["multi-domain learning","segmentation","universal model"],"search_terms":["net","universal","net","multi","domain","medical","image","segmentation","huang","han","yao","zhu","zhou"],"title":"3D U2-Net: A 3D Universal U-Net for Multi-domain Medical Image Segmentation","year":2019,"downloads":1}